基于矩阵信息几何样本筛选下局域联合处理的低空风切变风速估计方法  

Wind Speed Estimation of Low-Level Wind Shear Based on Matrix Information Geometric Sample Selection and Joint Domain Localized Processing

作　　者：李海[1] 李赞 LI Hai;LI Zan(Intelligent Signal and Image Processing Key Lab of Tianjin,Civil Aviation University of China,Tianjin 300300)

机构地区：[1]中国民航大学天津市智能信号与图像处理重点实验室,天津300300

出　　处：《火控雷达技术》2024年第2期1-8,共8页Fire Control Radar Technology

基　　金：国家重点研发计划项目(2021YFB1600600);天津市自然基金重点项目(20JCZDJC00490)。

摘　　要：针对非均匀杂波环境下,机载气象雷达低空风切变风速估计不准确的问题,本文提出了一种基于矩阵信息几何样本筛选下局域联合处理(Joint Domain Localized,JDL)的低空风切变风速估计方法。该方法首先提取雷达回波信号的特征信息建模为矩阵流形,然后通过选取几何度量方式估计几何均值矩阵,利用广义内积法(Generalized Inner Product,GIP)计算几何均值矩阵与各样本间相似度,并从训练样本中筛选样本相似度较高的训练样本估计待检测单元的杂波协方差矩阵,最后利用JDL方法构造降维变换矩阵,对杂波协方差矩阵进行降维处理,并构造降维处理器对非均匀杂波进行抑制,从而实现低空风切变风速估计。In this paper,a wind speed estimation method is proposed based on matrix information geometric sample selection and joint domain localized(JDL)processing for airborne weather radar to accurately estimate wind speed of low-level wind shear in non-uniform clutter environment.Firstly,the characteristic information of radar echo signals are extracted to model the signals as a matrix manifold.Then,geometric mean matrices are estimated by selecting geometric metric methods,and the similarity between geometric mean matrices and training samples are calculated using the generalized inner product(GIP)method.Samples with high similarity are selected to estimate the clutter covariance matrices for the units to be detected.Finally,dimensionality reduction of the clutter covariance matrices is performed through dimensionality reduction matrix constructed using the JDL method,based on which a dimensionality reduction processor is constructed to suppress non-uniform clutter,thus achieving wind speed estimation of low-level wind shear.

关 键 词：机载气象雷达 矩阵信息几何 低空风切变 风速估计 

分 类 号：TN959.4[电子电信—信号与信息处理]